‘2001: A Space Odyssey’ rendered in the style of Picasso
bhautikj.tumblr.com
bhautikj.tumblr.com
I can easily imagine an animation system like the one presented here enabling another massive improvement in animation efficiency. In the same way animation software allowed South Park to reuse pre-drawn objects, a deep learning system could enable south park to carefully define the entire drawing style just once, then generate complete episodes based on simple story boards and animation directives. Fortunately, South Park already has a significant amount of training data available, specifically every South Park episode yet produced.
I think at this moment it is not possible to instruct algorithm to take additional suggestions (artist ideas) into consideration when creating output image.
With vector graphics - where the lines and fills are mathematical objects - automatic 'tweening' becomes possible. Anime Studio (http://my.smithmicro.com/anime-studio-2D-animation-software....) is the zenith of this tech; there's also Synfig (http://www.synfig.org/cms/) and CACANi (https://cacani.sg/).
In Anime Studio it's possible to add all kinds of effects (including filter effects and motion blur) to animations, and to mix pure vector animation with cutout, or even frame-by-frame, animation.
Sometimes a perfect interpolation, or even something based on a physical model doesn't feel right, isn't what is expected.
This would be a great proof of concept, in fact someone should work on this as a pet project, but I doubt it would be worth as an investment in the rest of the show's run. I bet they don't spend a lot of time on the basic animation anymore, and instead focus mostly now on one-off set pieces and visual effects
http://www.comicbookscriptarchive.com/archive/panel-1/panel-...
On the first moon landing, quoted in The New York Times, (1969-07-21).
https://en.wikiquote.org/wiki/Pablo_Picasso
Curious about his feelings regarding this work. (I find it beautiful.)
I think it's superficial and doesn't do either source justice.
Also, for fairness, do the same with my comment:
"Unlike this status oriented nerd, I can relate to you and the author on a human level."
Referring to stepvhen's comment. I find it comical in all seriousness.
Even the must leisurely, casual or mundane topic, intended to be a refreshing change in colour (pun not intended considering the article but I'll take it) and\or conversation on Hacker News is subject to ridicule, assessment and biopsy.
It's why I like it here. Some of us are completely incapable of /not/ peer reviewing the shit out of everything. :)
Edit: words fail me, also spelt Stepvhen's handle wrong (sincerest apologies).
Then again I might take things too seriously. There was only one other comment when I posted, I didn't know what the tone would end up being.
Incidentally, I chose this handle because others always spelled my name "Steven", when it's "Stephen".
I never took your comment to not be polite. As mentioned, as a community. We're all quite guilty of being overly logical and serious. It's part of the charm.
I wasn't sure if I wanted to start collecting down votes for mentioning the Phteven meme (we don't like meme's here, do we?). But I had suspected that might be the basis for the spelling of your handle!
I'll down vote my own comments and see myself out...
> Then again I might take things too seriously.
if anything, too many people in this thread don't seem to be taking cubism and Picasso's style seriously either (signature faces? what ..?)
of course a deeplearning net can't actually do cubism.
if someone wrote a program to generate blocks of primary colours in pleasing ratios, really cleverly, it may look like Mondriaan to someone that has seen a couple of Mondriaan paintings. but everybody who knows what Mondriaan was trying to do, will instantly know that there's really no way today's computers could really perform the same process.
It's more like an insta-Picasso plug-in for one particular form of abstraction.
It's interesting and unusual, and yes, it would be better with constraints.
I'm not sure I'd want to look at it on a big screen though.
>of course a deeplearning net can't actually do cubism.
One of the interesting things to fall out of this research is the realisation that a lot of art - even figurative art - is based on abstraction of visual invariants.
There's no reason that creative abstraction can't be automated to create new styles.
The difference when humans do it is the level of psychological insight and feel for what's visually important and interesting in a scene.
That can probably be automated too, but it's a very much harder problem.
The challenge for most developers in this space is that they have a much more superficial understanding of art (and music, and writing) than they believe they do, so a lot of content and detail that's important to experienced viewers gets ignored. The result is superficial lookalike output - pastiche.
Technically, the superficial output is an achievement in itself, but it's still a way short of being artistically innovative in its own right.
It is also okay to pay homage to a great work of art, to sample it, to parody it, to outright copy it, to even forge it or pass it off as one's own so long as that is artistically done. However I feel this video is more like passing something through an electronic mangle than art. It is craft rather than art, even if it is hi-tech craft.
Had this technique been applied to an original short film that had its own footage and own way of telling an actual story then we could have had a winner.
2. I agree, as they are in this video, but I think cubism could be used to capture the experience of entering the monolith even better than Kubrick captured it.
I think the folly of the Cubists was to embrace the still life after the methods had been fleshed out. Most Cubist works are created from a handful of discrete images, this destroys the aspect of motion in my opinion. I've been working to capture proper motion using GoPro POV footage from various sources, rather then pausing the video I let the colours wing by and try and place them on the canvas. What I am choosing to perceive at the moment ends up represented (I avoid faces, thus I have floating forms without heads!), loose marks capture a door frame which has half of a tree-roadway. It's abstracted, but still based on a concrete sequence. The most effective are when a part of the frame is static throughout (a mounting bracket, bike, wiper blade, etc) because that imagery comes through only modified by changing light which provides a contrast against the jumble of forms and colours elsewhere.
The first step towards a Cubist film will involve multiple cameras shooting the same scene from multiple angles and recording the absolute time of the shot. [b]An algorithm that takes n images as an input and applies a deterministic transform to them (weighting certain images)[/b] trying to reconcile the overlaps, looking for prominent forms, picking/merging a colour to use, etc eventually mapping down to a single image. Assuming this could be developed the director would compose the scene from these multiple sources, perhaps giving a effect similar to [1]. The face of a lover is modified woven with images of a child and the image of meal prep on a sunny day. The conversation uses the word cheating in an innocent way and the focus snaps sharply back to the lover where scenes of cheating with the other/secretive behaviours/worry and paranoia are gently woven overtop eventually leaving the lovers voice as the only signifier of this conversation taking place.
But proper cubist imagery cannot come from a simple mapping of styles.
Also, as I've mentioned in my top-level comment, it's a great way to explore GPU programming and deep learning.
For those that haven't seen the movie, here's the trailer: https://www.youtube.com/watch?v=TY5PpGQ2OWY
It uses an algorithmic 'paint synthesizer' to generate brushes (with hundreds of presets) and auto-paint canvases, and is designed for animation (rotoscoping) as well as static artwork. The output can be reminiscent of the style of the movie 'A Scanner Darkly', but the software is hugely flexible. Here are a couple of rather amazing examples: http://studioartist.ning.com/video/auto-rotoscoped-dancers and http://studioartist.ning.com/video/dance-styles-animation-re...
Also, unlike most other 'painterly' software, the graphics are resolution independent - meaning that they can be scaled up to any size without loss of detail.
I think Alex J. Champandard's implementation is probably the best one out there right now. It has a ton of knobs to twist and is very fast.
Aside from the 'missing image' part, what is the fitness example here for training? How does the training process determine what a good image is given there aren't many (any?) examples of a source image -> picasso mapping?
I believe that this is done using Restricted Boltzman Machines[1] trained with the stylised image.
Think of it as a network that receives an image on the input layer, sends it to one or more hidden layers with less nodes (like an auto-encoder), and then tries to reconstruct the image on the output nodes. This is like a lossy compressor-decompressor overfitted to the stylized image.
Now, just pass the real image as an input to your network and the output should be a stylized version of the input.
It could be a way to exploit the mismatch of content and style as certain form of expression; but it may be more interesting if we can modify the temporal structure as well.
>Pretty tight that computers can drop acid now.
Anyway, here's a direct link to the video for mobile users: https://vimeo.com/169187915
There is no decomposing and superimposing points of views because, well, the data to do it it's just not there.
http://fairuse.stanford.edu/overview/fair-use/four-factors/#...
In a 1994 case, the Supreme Court emphasized this first factor as being a primary indicator of fair use. At issue is whether the material has been used to help create something new or merely copied verbatim into another work. When taking portions of copyrighted work, ask yourself the following questions:
Has the material you have taken from the original work been transformed by adding new expression or meaning? Was value added to the original by creating new information, new aesthetics, new insights, and understandings?
I think it's interesting that it's possible to create basically filters from existing images, but then applying those filters to large amounts of images (like in this movie) quickly loses the novelty effect and is just as boring as any photoshop or gimp filter became in the 90s after seeing it 3 times.
When I look at Picassos actual pictures, I am astonished and amazed with every new one I get to see. With these pictures, I get more and more bored with every additional image.
Watching Dave jog along the perimeter of Discovery One in perfect perspective sort of undermines the whole effect. Even though the images are painted over with Picasso-like textures, colors, and shapes, it doesn't really _look_ like the real thing. That said, even if I think it falls short of capturing exactly what makes a Picasso a Picasso, I still think it's pretty cool.
It needs some kind of averaging with nearby frames (or whatever), to avoid the constant flicker in areas of more or less solid color.
If you have a system with a recent-ish graphics card (I'm doing fine with my GTX 970), put a linux on it and check out the many GitHub projects that implement this stuff (some of the tools will only work on linux).
It's a great way to start learning about deep learning and GPU based computation , which are starting to look like very good things to have on your resume.
Plus, you get to make cool shit like this that you can actually show to your friends. I'm getting more interested in the text generation stuff as well too - I'd love to make a Trump speech generator :-)
If you pay attention to the video, the images are heavily upscaled.
Eventually with fast enough GPUs you could render a video game in this style, now that I would like to see.
Actually in the last year myriads of similar things were created, and this is simply boring.
This is as interesting as a random tumblr reblog. May be curious, but lacks any sense of achievement or originality.
John Holland, W. Brian Arthur, and Doyne Farmer (all associated with the Santa Fe Institute) have done much interesting work. Farmer talks of Theodor Wright, and his studies of aircraft manufacturing improvements during WWII, though much of Farmer's discussion seems to borrow heavily from Wright's browther Sowell, a geneticist looking at genetic drift and fitness landscapes. Quincy Wright's study of war also seems apropos. Interesting family.
INNOVATION!